Article Preview
Top2. Literature Survey
Pang et al., have considered sentiment classification as a special case of topic based on categorization aspect with positive and negative sentiments (Pang, Lee, & Vaithyanathan, 2002). They have undertaken the experiment with three standard algorithms i.e., Naive Bayes classification, Maximum Entropy classification and Support Vector machine being applied over the n-gram technique.
Pang and Lee have labeled sentences in the document as subjective or objective (Pang & Lee, 2004). They have applied machine learning classifier to the subjective group which prevents polarity classification from considering any misleading data. They have explored extraction of methods on the basis of minimum-cut formulation, which provides an effective way for integration of inter-sentence level information with bag of words.
Matsumoto et al., have considered the syntactic relationship among words as a basis of document level sentiment analysis (Matsumoto, Takamura, & Okumura, 2005). In this paper, frequent word sub-sequence and dependency sub-trees are extracted from sentences and they act as feature for SVM algorithm. They have used unigram method, bigram method and combination of both the methods for classification.
Read have proposed a source of training data based on language used in conjunction with emotions, could function independent of topic, domain and time (Read, 2005). Naive Bayes (NB) and Support Vector Machine (SVM) algorithms are applied to classify the polarity of the dataset.